INTERPRETATION OF MACROECONOMIC RESULTS FROM A CGE MODEL SUCH AS GTAP

Transcription

1 INTERPRETATION OF MACROECONOMIC RESULTS FROM A CGE MODEL SUCH AS GTAP Philip D. Adams Centre of Policy Studies, Monash University 1. Introduction This draft: 27 April 2003 There are four basic tasks in model-based analysis: 1. derivation of the model s theoretical structure; 2. calibration; 3. simulation design and solution; and 4. interpretation of results. This paper deals with the fourth task interpretation. Interpretation is concerned with explaining in a logical sequence the projections from a model. At each point in a properly-constructed sequence of explanation, the results are explained by drawing on results that have been explained earlier and/or values in the calibrated database and/or aspects of the underlying theory. Proper interpretation assists in the checking of the model s ilementation. It also enhances the credibility of the analysis and provides economic insights that are not obvious from a casual inspection of the results. The interpretation of results in terms of a logical sequence of connections is not an easy task in itself, and is often made more difficult by a requirement that economists not familiar with details of the model readily understand the interpretation. The objective of this paper is to outline a general strategy for interpreting macroeconomic results from a CGE model such as GTAP. 1 A general strategy is useful because it is available as a starting point and it is often suggestive of more detailed, simulation-specific interpretations. At the core of our general strategy is a stylised model of the model. 2 Stylised models assist both in identifying the principal theoretical mechanisms that underlie the projections from the full model and in highlighting iortant elements of the database. They can also be used to conduct sensitivity analysis, indicating the variability of projections from the main model to changes in the data and in other inputs. The rest of this paper is organised as follows. The general strategy is outlined in Section 2. The stylised macroeconomic model is derived in Section 3. In Section 4, the power of the general 1 Our strategy focuses on macroeconomic variables rather than on detailed structural variables such as industry output. Typically, consideration of the macro picture leads to a better understanding of the micro detail. For exale, understanding what happens to aggregate private consution is a key to understanding the response of consution-oriented industries such as dwelling ownership. Similarly, understanding what happens to the real exchange rate is the key to understanding the response of trade-exposed industries. 2 Stylised models are used extensively at the Centre of Policy Studies to facilitate "back-of-the-envelope" explanations. A recent exale is Dixon and Rimmer (2002, Section 7.2). The usefulness of stylised models and "back-of-the-envelope" explanations is discussed in detail in Dixon, Parmenter and Powell (1984). 1

2 strategy is demonstrated with an application to a GTAP simulation of the effects of APEC trade liberalisation. Concluding remarks are in Section A General Strategy Experience suggests that there are two keys to a good strategy for macroeconomic interpretation. The first is to identify the first point of iact of the exogenous shock. This will generally suggest a way to enter an otherwise closed loop of results. The second is to use a stylised model to explain the outcomes for factor quantities and for real factor prices. From these, explanations normally follow for the major coonents of real from the income and expenditure sides, and for other macroeconomic variables such as the real exchange rate, the deflator and the balance on trade account. 2.1 Identifying first-round iacts We set out three exales of simulations in which interpretation of the results most naturally begins with the initial points of iact. The first exale is a simulation of the effects of a unilateral cut in one country s tariff protection against iorts (the focus of our worked exale in Section 4). This reduces the purchasers price of iorted goods relative to the purchasers price of domestic substitutes, and increases demand for iorts at the expense of demand for domestic substitutes. It can also be shown that the tariff cut reduces the real cost of fixed factors 3. This is the starting point of our detailed investigation in Section 4. Nothing other than the exogenously iosed change, the underlying theory of the model and the data to which the theory is calibrated explains the change in relative demand and the initial reduction in the real cost of the fixed factor. A second exale is a simulation of the iact of a tax on greenhouse gas emissions. Iosition of such a tax has two obvious first-round effects. First it makes greenhouse-gasintensive products such as coal-generated electricity more expensive relative to cleaner alternatives such as gas-generated electricity. Second it provides a new source of government revenue. Either is an entry point in the eventual macroeconomic interpretation. The final exale is a simulation of an increase in the efficiency of use of all primary factors in an economy. The immediate macroeconomic iact of this is an increase in real. Real increases by the full amount of the cost savings arising from the iroved efficiency. Another way to look at it is that the real return to the fixed factor(s) increases because the fixed factor(s) receive all of the benefits of the iroved efficiency. These are the entry points in the macroeconomic interpretation of the shock. 2.2 Using a stylised model The next step in our general methodology is to use a stylised model to explain the outcomes for macroeconomic indicators. Formalised within the stylised model should be aspects of the macro closure adopted in the full model. 4 3 Typical coarative-static closures have one of labour or capital fixed, but not both. For exale, in a long-run closure labour is typically fixed and capital is variable. With labour fixed the real wage rate can vary. With Capital variable, the rate of return on capital is fixed. 4 Understanding the closure is iortant in understanding any set of results. For exale, in the standard GTAP closure the quantities of all endowed inputs (labour, capital and agricultural land) in each region are exogenous and are typically set to zero. This has 2

3 The stylised macroeconomic model developed in Section 3 has a multi-country dimension with each variable and coefficient having a regional index. However, there is no explicit modelling of the linkages between each of the economies arising from trade in goods and services and from international financial flows. Instead, just as in a single-country model, each economy is modelled in isolation, with allowance for exogenous changes in the positions of foreign-demand schedules for exports and foreign-supply schedules for iorts as well as for changes in the global rate of return on capital. We adopt the single-country approach for the sake of silicity. A fully endogenous explanation of trade in goods, services and financial instruments is colex and adds relatively little to the insights that can be gained from a model that excludes them. 3. A Stylised Macroeconomic Model Equations of the stylised model are shown in Table 1. Equation (1) defines real at market prices in region r ( Y ) as the sum of real consution (C), real investment (I), real government consution (G) and the net volume of trade (X-M). 5 Equation (2) is region r s production function. It relates real at factor cost ( ) to inputs of labour (L) and capital (K), with an allowance for changes in all-factor technical efficiency. A is the technological-change term. An increase in A ilies technological deterioration. A decrease ilies technological irovement. In writing (2), and elsewhere in the stylised model, we ignore the existence of agricultural land. The relationship between real at market prices and real at factor cost is given in equation (3). Y TAX is a quantity index, which represents the quantity on which indirect taxes are applied. In other words, TAX Tax Y =, FC T P where: Tax is the economy-wide collection of indirect taxes net of subsidies; T is the average ad valorem rate of indirect taxes less subsidies; and P FC is the price of at factor cost (assumed to be the price to which T is applied). Note that for the remainder of this paper we assume that iort taxes are the only form of indirect taxation iosed in each region. Thus T is taken to be the average ad valorem rate of iort tariff applied in region r. Equation (4) is the economy s consution function, relating the value of private consution ( P C C ) to the value of at market prices via the average propensity to consume ( Ω ). This explanation of private consution is consistent with GTAP s treatment of the regional household. Equation (5) explains the value of government consution. It is analogous to equation (4), with Γ being the government s average propensity to consume. Equation (6) relates the volume of iorts to the general level of activity ( Y ), to the real exchange rate (RER) and (negatively) to the power of the average rate of tariff, (1+T). An increase in the real exchange rate (a real appreciation) with all else unchanged signals a decline in coetitiveness of traded-goods industries in region r and hence an increase in iorts. Y FC strong ilications for the determination of real and for relative factor prices. It constrains the response and it forces real factor prices to move together. 5 at factor cost is the total cost of primary factors. It is equal to at market prices less the value of indirect taxes net of subsidies. 3

4 Equation (7) is region r s export equation. Exports are inversely related to the real exchange rate and to Y W, an exogenous variable representing the general level of activity in the countries to which region r sells. Equation (8) determines economy r s investment, by setting the ratio of investment to capital in the solution year to the exogenous variable Φ. Equation (9) defines the real exchange rate as the ratio of the price of in the domestic economy (a proxy for local production costs) to the average price of in region r s trading partners ( P W ), converted to domestic currency via the exchange rate Θ ). The exchange rate is expressed as domestic currency per unit of foreign currency. (10). The price of at market prices is related to the price of at factor cost via equation Equation (11) relates the terms of trade (TOT) (i.e., the price of exports relative to the price of iorts) negatively to the volume of exports and to the average price of iorts paid by region r ( P W ). F TOT {} is a positive increasing function of X. Equation (11) is consistent with region r being a small country with respect to iorts, but facing a downward-sloping demand curve for its exports. Equation (12) shows the relationship between the terms of trade and the ratio of the price of private consution to the price of at market prices. F P {} is a positive increasing function of C C TOT. Changes in ( P / P ) will often be associated with changes in the terms of trade. A terms-of-trade irovement reduces the price of total domestic final expenditure (which includes iorts but not exports) relative to the market price of output (which includes exports but not 6 iorts). Equation (13) explains the ratio of the price of government consution to the price of in a way that is analogous to the explanation of the price of private consution relative to the price of in equation (12). Equation (14) connects relative factor inputs to relative factor prices. RP L is the real price of labour, defined as the nominal wage rate relative to the price of at factor cost (the price of output). RP K is the real price of capital, defined as the nominal rental on capital relative to the price of at factor cost. Under (14), an increase in the real price of labour relative to the real price of capital will cause an increase in the capital intensity of the economy. Notice that with perfect coetition, the real price of labour is equivalent to the marginal product of labour and that the real price of capital is equivalent to the marginal product of capital. The relationship between real factor prices is given in equation (15). It is derived from the geometric-average form: FC S L K 7 P = (P A) (P A) (18), L K S where S L and S K reflect the shares of labour and capital in the economy (SL ( r) + SK = 1), and P L and P K are the nominal wage rate and the nominal rental on capital. To obtain (15), 6 Note, that iorts of private consution goods corise only a fraction of total iorts and so not all terms-of-trade changes will lead to changes in the price of private consution goods relative to the deflator. Moreover, private consution is only one part of total domestic final expenditure (GNE). Consequently, changes in the price of labour, say, may affect the prices of other GNE coonents more than the price of private consution, leading to a change in the price of consution relative to the deflator even when there has been no change in the terms of trade. 7 At this point the equation numbering might appear confusing. This is because the equations of the model listed in Table 1 are numbered consecutively and the equations in the text follow on from the number of the final model equation, namely equation number 17. 4

5 divide both sides of (18) by the factor-cost price of, redefine in terms of real prices, then rearrange. Equation (16) explains the real price of labour as a function of the real wage rate (RW) (i.e., the nominal wage rate deflated by the price of consution), the inverse of the terms of trade, and one plus the ad valorem rate of indirect taxes less subsidies. In deriving (16), we note that the real price of labour equals: C PL PL P P = (19). FC C FC P P P P C On the right hand side of (19), PL / P is the real wage rate (RW), P / P is a FC function of the inverse of the terms of trade (see equation (12)), and P / P equals (1+T) (see equation (10)). P P K FC In deriving (17), note that the real rental price of capital equals: I PK P P = (20), I FC P P P where P I is the price of investment. On the right-hand-side of (20), PK / P can be interpreted as the economy's rate of return, ROR. The second term, like the corresponding term in 8 (19), responds to changes in the inverse of the terms of trade. The final term is equivalent to (1+T). The linearised percentage-change forms of equations (1) to (17) are given in the second part of Table 1 as equations (1 ) to (17 ). Lower-case letters identify percentage changes in variables written in the corresponding upper-case letters. Coefficients and parameters in the linearised model are given in Table 2. The derivations of the linear forms are generally straightforward but some explanation is necessary. Equation (6 ) is an ilication of the CES form for domestic/iort substitution, with σ M representing the average domestic/iort substitution elasticity. Omitting the regional index for sake of brevity, note that the percentage change in (1+T) is: 100 T 100 T = (1 + T) (1 + T) T T T T P = t = t (1 + T) (1 + T) FC FC Y FC FC PY where, as noted in Table 2, S T is the share of indirect taxes net of subsidies in at market prices. In equation (7 ) and equation (11 ), σ X (r ) is a negative parameter, representing the price elasticity of world demand for exports. Equation (11 ) also contains a shift variable, f X, that allows for exogenous shifts in the foreign-currency price of exports in region r. = ts T (21), For equation (12'), notice that from the expenditure side of we can write: c g pgdp = SC p + SG p + SI p + (SX p SM p i x C m I ). To obtain (12'), we assume that the prices of private consution, public consution and investment move together, i.e., 8 Footnote 6, which relates to consution goods, applies in an analogous way to investment goods. 5

6 c g p = p = p, i and that trade is initially balanced, i.e., Thus, SX = S M c. x pgdp = p + SX (p p ) = p + SX tot m c (22). In equation (13 ), η is an exogenous shift variable that allows for changes in the ratio of the price of government consution to the price of private consution. Equation (14') is an ilication of the CES form for capital/labour substitution: K L σkl FC PL P = FC PK P (23), where σ KL is a positive parameter, representing the capital/labour substitution elasticity. Our starting point for equation (16') is equation (19). Equation (19) ilies that the percentage change in real price of labour is the sum of: (a) the percentage change in the real wage rate (rw); (b) the percentage change in the ratio of the consution price to the market-price deflator; and (c) the percentage change in the ratio of the market-price deflator to the factor-cost deflator. Coonent (b) is explained by equation (12') and coonent (c) is explained by equation (10'). The derivation of equation (17') from equation (20) is analogous to the derivation of (16') from (19). Table 3 lists the variables in the linearised form of the stylised macro model. The number of variables (30) exceeds the number of equations (17) by thirteen. Thus, thirteen variables must be set exogenously. Table 3 shows the exogenous/endogenous status of each variable that has a natural classification. Eleven variables are shown as naturally exogenous; in general, the model cannot explain changes in these variables. The nominal exchange rate is one such variable. It acts as the numeraire, namely, it determines the general price level. 9 This leaves two more variables to be made exogenous. In short-run coarative-static simulations, the capital stock and real wage rate are typically exogenous, allowing the model to determine values for eloyment and the rate of return on capital. In long-run coarative-static simulations, eloyment and the rate of return are exogenous, allowing the model to determine values for capital and the real wage rate. 4. Interpretation: GTAP projections of the effects of APEC liberalisation 4.1 GTAP Simulation The version of GTAP used here is that documented in Hertel (1996) with some additional equations and variables described in Appendix A of Adams, et al. (1997). Calibration was based on data from the version 5 database (Dimaranan and McDougall, 2002). The solution software was GEACK (Harrison and Pearson, 1996). Table 4 shows the regions distinguished. The industry classification, while not of iortance for this study, is shown in Table 5 for the sake of coleteness. 9 The choice of numeraire is arbitrary. Natural alternatives to the exchange rate as the numeraire are the price of consution or one of the deflators. 6

7 Simulation design The GTAP simulation involved the removal in each APEC country of all ad valorem iort tariffs as well as tariff equivalents of bilateral non-tariff barriers on APEC-sourced iorts. 10 Iort protection between the rest of the world (ROW) and APEC is maintained. Table 6 shows the percentage deviations in the average tariff rate of each region (first row of numbers). The deviations due to the removal of tariffs were couted under the following assutions about factor markets and about macro behaviour in regions. capital is mobile, moving across regions to equalize disturbances in rates of return generated by the tariff shocks ; aggregate eloyment of labour and of land is fixed in each region; government budget balances are slack 11 ; and in the solution year, investment and capital in each region move together, with the world rate of return adjusting to ensure that the weighted sum of changes in each region s investment equals the change in global savings. Because of the capital-mobility assution, we describe our simulation as long run. 12 Table 7 shows initial values for the coefficients identified in Table 3 derived from the version 5 GTAP database. Table 8 shows results for the 17 variables identified as endogenous (long-run closure) in Table 3. The GTAP projections are shown in the columns labelled GTAP. Overview of GTAP projections As shown in the first row of numbers in Table 8, the real- gains from APEC-trade liberalisation tend to favour regions that are small and open. Exales are Thailand/Philippines (THA/PHL), for which an 18.2 per cent gain in real is projected, and Malaysia/Singapore (MYS/SGP), for which an 8.2 per cent gain is projected. Apart from being small and open, these regions also have high pre-liberalisation levels of protection against iorts, especially against iorts of investment goods. The regions that gain least in terms of real are the ROW, which does not cut protection and the large APEC economies, North America (NAM) and Japan (JPN). Projections for changes in the terms of trade as a result of the APEC liberalisation are given in row 10 of Table 8 ( GTAP column). Here, the liberalisation favours JPN, the agricultural exporting regions such as Australia (AUS) and New Zealand (NZL), and Taiwan (TWN). JPN s terms of trade iroves, in part, because of strong increases in NAM demand for Japanese products due to tariff liberalisation. The terms-of-trade increases in AUS and NZL flow from the existence, prior to liberalisation, of large barriers iosed on agricultural and food commodities from these countries. The terms-of-trade increase in TWN stems mainly from strong increases in demand in other APEC regions for iorts of other machinery and equipment from TWN. Regions that experience terms-of trade losses as a result of APEC liberalisation do so either because they are 10 From now on, whenever we use the term tariff we mean both tariffs and tariff equivalents of other trade barriers. 11 This ilies that the deterioration in government budget balances caused by the loss of tariff revenue is not offset by reduced government spending or by increases in other taxes. Most published GTAP applications adopt this assution. We have done so for convenience, although it is clearly incorrect. 12 GTAP allows other assutions than those adopted here with respect to factor markets and macro behaviour. Each set of assutions describes a closure of the model. The standard GTAP closure is a short-run closure in which each region s endowments of capital, labour and land are held fixed. 7

8 agricultural iorting regions or because the ROW (where real activity falls because of liberalisation) is over represented as a destination for their exports. Rows 5 and 6 of Table 8 ( GTAP columns) show the trade-enhancing effects in each APEC region of tariff reductions. All APEC regions are projected to experience increases in trade volumes, with the largest increases occurring in regions with the highest initial levels of protection. Rows 2, 3 and 4 of Table 8 ( GTAP columns) show the iact of liberalisation on the coonents of real domestic demand. Aggregate real consution (private plus government) is elevated in all APEC regions other than NAM and JPN. The largest increases are in regions that are projected to experience the largest gains in real. 4.2 More detailed interpretation aided by the stylised model The overview above is fairly typical of explanations offered in published journal articles of results from a GTAP simulation. Even so, in the context of the APEC liberalisation simulation, the overview leaves many questions unanswered. For exale, why did real in NAM fall? What accounts for the changes in factor prices (changes that are so iortant for income distribution analysis)? Why did NZL gain by more than AUS? What accounts for the large gains in THA/PHL, and for the relatively small gain in Indonesia (IDN)? Answers to these and to many other questions, are revealed with the aid of the stylised macro model. There are two ways that the stylised model can be used. One is to use individual equations in isolation to understand specific results from the full model. For exale, we might use equation (2 ) to explain outcomes for real at factor cost based on GTAP values for the righthand coefficients and GTAP-simulated values for right-hand variables. The second way that the stylised model can be used is to eloy it as a model in its own right to reproduce the simulation of the primary model using shocks aggregated to a macroeconomic level. Results from the stylised simulation can then be coared with those from the simulation of the primary model, with special attention directed towards significant variations. We adopt this second approach in the discussion below. Simulation design for the stylised model The GTAP simulation of the long-run effects of APEC trade liberalisation is reproduced in the stylised macro model by shocking t and variables that are naturally exogenous in the stylised macro model but are naturally endogenous in the GTAP model, namely y W, y TAX, p W and ror. The shocks to t are equivalent to the shocks iosed in the GTAP simulation (see TAX Table 6). The shocks to y W and y ( r) were calibrated so that the stylised model reproduced exactly the GTAP projections for the volume of exports (x in the stylised model) and the terms of trade (tot in the stylised model) (see Table 8). y TAX was set equal to the value of m, reflecting the fact that iorts are the tax base for the taxes that are being removed. The shocks to ror were uniform and set equal to the GTAP outcome for the global rate of return on capital (rorg), This is the increase necessary to ensure that the change in global investment matches the change in global savings arising from APEC liberalisation. Detailed explanation The projections from the stylised macro model are given in Table 8 in the columns labelled Stylised. Our explanation of the GTAP projections using the stylised mechanism starts with real 8

9 factor prices and the use of factor inputs. It then proceeds to real, to the price of and to the expenditure-side coonents of real. Changes in real factor prices Equation (17 ) shows that there are three main mechanisms through which tariff cuts can affect the real cost of capital (the nominal cost of capital relative to the deflator at factor cost). The first is via a change in the global rate of return on capital essentially the rate of return required on the global market for capital. The second is via the direct effects of the tariff cuts on the duty-paid prices of iorted inputs to investment. The third is via changes in the terms-of-trade that affect the average c.i.f. price of iorted capital goods relative to the deflator. In the GTAP simulation, the rate of return on capital increases in all regions by 1.28 per cent, whereas the outcomes for the terms of trade vary across regions. Ultimately, the GTAP-simulated value for rp k is negative for all regions except NAM, JPN and ROW. In the exceptional cases, the iact of the increase in the required rate of return on the real cost of capital more than offsets the effects of changes in the terms of trade and in tariff rates. In the GTAP simulation, the regions that experience the largest falls in the real cost of capital are THA/PHL, TWN, and CHINA. The tariff cuts as a share of in these regions are relatively large (see the second row of Table 6) and dominate the influences of other factors on the cost of capital. Table 6 shows that the tariff cuts in MYS/SGP are also large, but the GTAPprojected fall in that region s real capital cost is mild relative to the fall projected for THA/PHL. To understand this result, note that the stylised model s projection for MYS/SGP is relatively close to the GTAP projection, whereas for THA/PHL, TWN and CHINA the stylised macro model underestimates the fall projected by GTAP. This points to coositional effects that are not accounted for in the stylised model. Closer inspection of the GTAP projections shows that the tariff cuts in THA/PHL, TWN and CHINA are more oriented towards investment goods than are the cuts in MYS/SGP. Thus in the most-affected regions the falls in the price of investment relative to the price of (market prices) are much larger than in MYS/SGP. This is why the tariff cuts in the most-affected regions have a stronger influence on the real cost of capital than they do in MYS/SGP. Equation (15 ) establishes that there is a strong negative relationship between the real cost of labour and the real cost of capital. The degree of responsiveness is determined by values for the ratio of the share of capital to the share of labour, initial estimates of which can be deduced using data in Table 7 for S K S L. These show that the responsiveness of the real cost of labour to a one per cent change in the real cost of capital will be greatest in the developing regions such as THA/PHL. In these regions wages are low and so the ratios of capital payments to labour payments are high. In developed regions, the situation is reversed; coaratively high wage rates means that the ratio of capital payments to labour payments is low. The negative relationship between real factor prices is confirmed by the GTAP results in Table 8. In regions where the real cost of capital declines the real cost of labour rises. The largest rise occurs in THA/PHL, which is projected to experience the largest fall in the real cost of capital and which has a relatively high capital share. For some regions the stylised macro model fails to account adequately for changes in the real cost of land as projected by GTAP. This is due primarily to the omission of agricultural land. In GTAP, there are two main fixed factors of production in the long-run, labour and land. Thus when the real cost of capital falls, the real cost of labour plus land rises. The distribution of this increase 9

10 in each region between labour and land depends partly on the respective shares of labour and land in. In agriculturally-oriented regions like AUS and NZL, in which land s -share is over 2 per cent, land absorbs a significant portion of the benefit that would otherwise have gone to labour. In these cases, the stylised macro model will tend to overestimate the increase in the real price of labour. Changes in factor inputs and real at factor cost Equation (14 ) shows that at the macroeconomic level a fall in the real cost of capital relative to the real cost of labour causes producers to substitute capital for labour..13 Accordingly, with eloyment unchanged in the long-run, capital is created in regions where the real cost of capital falls (see rows 8 and 16 of Table 8) and is destroyed in regions where the real cost of capital rises. THA/PHL experiences the largest increase in capital, followed by CHINA, TWN and SKOR. ROW and NAM experience the largest falls in capital. Having explained changes in capital stocks we have also explained changes in real (see equation (2 )). 14 Figure 1 is a chart of the deviations in real at factor cost as projected by GTAP and as projected by the stylised macro model. This figure illustrates two things: the relativities across regions in factor-cost responses, and the reliability of the stylised model s explanation of deviations in real factor-cost. Real at factor cost falls in three regions, NAM, JPN and ROW. These falls ultimately reflect projected increases in the real cost of capital in each adversely-affected region arising from the global increase in the rate of return on capital. Liberalisation is projected to increase real in all other regions. The regions that expand least are AUS, NZL and IDN. This is consistent with the relatively small tariff cuts in these regions (see Table 6). Note that NZL gains more than AUS despite a smaller tariff cut because the NZL economy is more capital intensive and thus more responsive in a long-run simulation. Figure 1 shows that the stylised model underestimates the true effects on real factor-cost most significantly for THA/PHL, TWN, AUS and IDN, and overestimates the effects most significantly in CHINA. Underestimation can generally be traced back to an underestimation of the reduction in the real cost of capital. In all of the cases noted above, the error in estimating the iact on the real cost of capital is accounted for by the coosition of the tariff shock which is disproportionately weighted towards iorted investment goods (see the discussion regarding real factor prices above). Overestimation of the factor-cost result for CHINA, though, does not stem from an overestimation of the iact on the real price of capital. Indeed, as can be seen in Table 8 the stylised model underestimates the iact on the cost of capital. The key problem here appears to the relationship between the real cost of capital and the real cost of labour. With the stylised model, a 4.2 per cent reduction in the real price of capital translates into a 5.5 per cent increase in the real price of labour. By contrast, in the GTAP simulation a 5.3 per cent increase in the real cost of capital results in only a 3.3 per cent increase in the real price of labour. Again, the omission of land is the main cause of the disagreement. Though the share of land in in CHINA is initially quite 13 In a multi-sectoral model, the economy-wide ratio of capital to labour is affected not only by changes in the ratio of aggregate factor prices but also by changes in the sectoral coosition of. For exale, in the long-run the capital to labour ratio can rise because of changes in the coosition of the economy towards capital-intensive industries. Conversely, it can fall with changes in coosition away from capital-intensive industries. 14 Recall that in the long-run simulation eloyment is fixed and there is no change in industry technologies. 10

11 small, the increase in land rental in CHINA as projected by GTAP is very large (69 per cent), displacing a significant portion of the benefit that otherwise would have gone to labour. The price of Chinese agricultural land rises strongly because of the significant level of pre-liberalisation tariffs iosed on Chinese agricultural iorts by other APEC regions. Changes in real at market prices and in the expenditure-side aggregates of real Real at market prices is related to real at factor cost by the addition of real indirect taxes net of subsidies (see equation (3)). In the simulation of the stylised macro model we assume that real indirect taxes net of subsidies move with the volume of iorts. Accordingly, in the stylised simulation the ratio of real at market prices to real at factor cost (row 1 less row 7 in Table 8) increases in all regions where iorts rise. This is consistent with the GTAP projections. Note, though, that the stylised model typically overestimates the increase in -ratio. This is because the stimulation of iorts does not reflect the direct iact of tariff cuts only. Some of the stimulation reflects general expansion in final demand and so the percentage increase in iort volume will typically overstate the GTAP-projected percentage increase in real indirect taxes net of subsidies. Figure 1: Long-run deviations in real at factor cost percentage deviation GTAP Stylised -5.0 NAM JPN AUS NZL CHINA SKOR TWN MYS/SGP THA/PHL IDN ROW We now turn to the expenditure-side aggregates of real. In the stylised macro model, the change in real private consution is explained by equation (4'). Assuming no change in the average propensity to consume ( Ω ), and no change in the ratio of the price of at market prices to the price of private consution the stylised model would predict that real consution moves in line with real at market prices. Inspection of the GTAP results shows that this is a reasonable approximation (coare rows 1 and 2 in Table 8). Changes in the ratio of the price of to the price of household consution, though, are iortant in some regions (coare row 9 with row 11). In the stylised model, changes in this price ratio are linked to changes in the terms of trade (see equation 12 ). According to equation (12 ), in regions where the terms of trade irove, the ratio of the price of to the price of consution increases and allows room for an increase in real consution. Overall, the stylised model provides an adequate approximation to the iacts on 11

12 private consution of tariff liberalisation, taking into account errors in the estimation of the percentage change in. The stylised model s account of changes in real public consution is less convincing than its account of changes in real private consution. This is so because in the GTAP simulation the price of public consution does not move with the price of private consution, as is assumed in equation (13 ). In this case coositional effects are iortant. Private consution tends to be oriented towards capital-intensive goods (housing, for exale), with a sizeable share of iorts in total expenditure. Public consution on the other hand is oriented towards labour-intensive goods (such as public administration and community services), with only a small share of iorts in total expenditure. It follows that in simulations where wage and iort-price effects are significant, the price of public expenditure can move quite differently from the price of private consution. This is certainly the case in the GTAP projections for SKOR. Investment in both the stylised macro model and GTAP is indexed to capital in long-run simulations. Projections for capital are discussed above. We turn next to trade volumes. Equation (6 ) suggests that in the GTAP simulation changes in iort volumes reflect three factors: (a) the changes in relative domestic/iort prices due to the direct effects of the tariff cuts; (b) changes in domestic demand; and (c) changes in the real exchange rate. In the GTAP simulation, the iact on iorts of (a) is unambiguously positive. In most regions, the iort-intensive coonents of demand are investment-related. Thus in most of the APEC regions the effects of (b) reinforce the effects of (a). The final factor is changes in real exchange rates. In the stylised macro model, the change in real (Y) in each region is determined, in the main, by changes in relative factor prices, and the change in real domestic absorption (C+I+G) in each region reflects the change in real. It follows that the change in net trade volume (X-M) is constrained by the requirement for macroeconomic balance (Y=C+I+G+X-M)). In other words the volume of net trade has to adjust to reconcile changes in real with changes in real domestic absorption. The residual adjustments occur via changes in real exchange rates. 15 In the simulation of the stylised macro model, the change in real domestic absorption in each region generally exceeds that of real because C and G move with, but I moves with capital. This means that the balance of trade for each APEC region will deteriorate in real terms. The stimuli to iort volumes from effects (a) and (b) generate some of the deterioration in each economy. The residual adjustment occurs via changes in real exchange rates. Real appreciation is required in most cases (see row 13 of Table 8). However, in some cases (e.g. THA/PHL) where effects (a) and (b) are very large, real devaluation is necessary. The pattern of devaluation and appreciation suggested by the stylised model is consistent with that projected by GTAP. 5. Conclusion In this paper we suggest a general strategy for interpretation of macroeconomic projections from a CGE model like GTAP. Central to this strategy is the use of a stylised macro model. Experience suggests that the stylised framework provides a robust method for understanding CGE 15 Appreciation of the real exchange rate reduces the coetitiveness of traded-goods industries versus foreign substitutes. This leads to fewer exports and to more iorts. Conversely, devaluation of the real exchange rate enhances coetitiveness, leading to more exports and to fewer iorts. 12

13 projections. In this paper, the robustness is put to the test with an analysis of GTAP projections of the long-run iacts of APEC trade liberalisation. The stylised model assists in identifying the principal theoretical mechanisms that underlie the projections from the full model. It also aids in highlighting the iortant elements of the database (see Table 7). In the context of the long-run APEC liberalisation simulation, insights gained from the stylised model include the following. 1. Changes in relative factor prices play an iortant role in determining the response in a fixed-factor simulation. 2. Tariff cuts that are oriented towards investment goods have a larger positive iact on real in a long-run simulation, all else unchanged, than tariff cuts oriented towards labour-intensive goods Adjustment of land rentals is an iortant consideration in understanding the macro response to a tariff cut of agricultural-dependent countries. 4. The size of the response is positively related to the size of the tariff cut as a share of. 5. Changes in the real exchange rate reflect changes in real relative to changes in real domestic absorption. 6. Changes in the terms of trade play an iortant role in determining the size of the response and in determining the coosition of that response across the large expenditure-side aggregates. References Adams, P.D., K.M. Huff, R. McDougall, R. Pearson and A.A. Powell (1997), Medium- and Long- Run Consequences for Australia of an APEC Free-Trade Area: CGE Analyses Using the GTAP and Monash Models, Asia-Pacific Economic Review, 3(1), pages Dimaranan, B.V. and R.A. McDougall (2002). Global Trade, Assistance, and Production: The GTAP 5 Data Base, Center for Global Trade Analysis, Purdue University. Dixon, P.B., B.R. Parmenter and A.A. Powell (1984), "The Role of Miniatures in Coutable General Equilibrium Modelling: Experience from ORANI", Economic Modelling, Vol. 1, No. 4, pages Dixon, P.B. and M.T. Rimmer (2002), Dynamic General Equilibrium Modelling for Forecasting and Policy: A Practical Guide and Documentation of MONASH, North-Holland Publishing Coany, Amsterdam. Harrison, W.J. and Pearson, K.R. (1996), Couting Solutions for large General Equilibrium Models using GEACK, Coutational Economics, Vol. 9, pp Hertel, T.W. (ed.) (1996), Global Trade Analysis: Modelling and Applications, Cambridge University Press, Cambridge. 16 The reverse is true in a short-run simulation, with labour variable and capital fixed. 13

14 Table 1: A Stylised Macro Model Useful for Analysing Results from MMRF-Green Levels equations Y = C + I + G + (X M) (1) Y Y P P FC C G A = F = Y FC Y + Y C = Ω P G = Γ P (L,K) TAX (2) (3) Y Y M = FM (Y,RER,1/(1 + T)) X I / K RER = F ( RER) Y (7) X W = Φ (8) = P /( Θ P W FC P = P (1 + T) TOT P P C G / P / P ) ( X ) P } (4) (5) (6) (9) (10) = 1/{F (11) K / L S TOT = 1/ F = 1/ F P C P G W (TOT) (TOT) (12) (13) = FKL ( RPL / RPK ) (14) S L K RP = RP A (15) L K RP = F (RW,(1/ TOT),(1 T) (16) RP L RP L + = F (ROR,(1/ TOT),(1 T)) (17) K RP K + 14

15 Table 1 (continued): A Stylised Macro Model Useful for Analysing Results from MMRF-Green Linearised equations in the percentage changes of variables y y fc = S = S C L c + S i + S l + S I K (S X G g + x S fc tax y = SFC y + ST y c = ω + (p g = γ + (p m x i rer = y gdp gdp + σ + y + y M = σx rer + y M m) (1 ) k a (2 ) ) p ) p (rer S W T c g (3 ) (4 ) (5 ) t) (6 ) (7 ) k = φ (8 ) = pgdp ( θ + p fc W ) (9 ) pgdp = pgdp + ST t (10 ) tot c p = p p g k rp rp rp = (1/ σ ) x + f p (11 ) = p gdp c X S X X W tot (12') + η (13') l = σ (rp rp ) (14 ) KL L K = ( S / S ) rp a (15 ) L K L K + = rw S tot + S t (16 ) L X T = ror S tot + S t (17 ) K X T 15

16 Table 2: Coefficients and Parameters in the Percentage-change form of the Stylised Model Symbol S C S I S G S X S M S L S K S FC S T σm σx σkl Description Share of private consution in at market prices Share of investment in at market prices Share of public consution in at market prices Share of exports in at market prices Share of iorts in at market prices Share of the cost of labour in at factor cost Share of the cost of capital in at factor cost Share of the cost of factors (labour and capital) in market-price Share of indirect taxes less subsidies in market-price Average domestic/iorted substitution elasticity. Price elasticity of world demand for region r s exports (negative parameter) Capital/labour substitution elasticity 16

17 Table 3: Percentage-change Variables of the Stylised Model Symbol Description Naturally exogenous(x)/ endogenous (N) Shortrun Longrun y Real at market prices N c Real private consution N i Real investment N g Real public consution N x Real exports N m Real iorts N y fc Real at factor cost N l Eloyment N X k Capital stock X N a All factor technical efficiency X y tax Quantity to which indirect taxes are attached. X ω Ratio of private consution to X γ Ratio of public consution to X p gdp price (market prices) N θ Nominal exchange rate (domestic per foreign) X (numeraire) tot Terms of trade N p w Average foreign-currency price of iorts X f X Shift in foreign-currency price of exports X p c Price of private consution N p g Price of government consution N η Ratio government to private consution prices X rer Real exchange rate N y W Average change - region r s trading partners. X ror Rate of return on capital N X φ Ratio of investment to capital in solution year X t Ad valorem rate of indirect taxes X p fc gdp price (factor cost) N rp l Real price of labour N rp k Real price of capital N rw Real wage rate X N 17

18 Table 4: Regional aggregation of GTAP used for this paper Identifier Countries in region 1. NAM North America (United States, Canada and Mexico) 2. JPN Japan 3. AUS Australia 4. NZL New Zealand 5. CHINA China (includes Hong Kong) 6. SKOR South Korea 7. TWN Taiwan 8. MYS_SGP Malaysia and Singapore 9. THA_PHL Thailand and the Philippines 10. IDN Indonesia 11. ROW Rest of the World Table 5: Sectoral aggregation of GTAP used for this paper 1.PDR Paddy Rice 20.LEA Leather, etc. 2.WHT Wheat 21.LUM Lumber 3.GRO Grains 22.PPP Pulp, Paper, etc. 4.NGC Non-Grains Crops 23.P_C Petroleum and Coal 5.WOL Wool 24.CRP Chemicals, Rubbers, Plastics 6.OLP Other Livestock 25.NMM Non-Metallic Minerals 7.FOR Forestry 26.I_S Primary Ferrous Metals 8.FSH Fisheries 27.NFM Non-Ferrous Metals 9.COL Coal 28.F Fabricated Metal Products 10.OIL Oil 29.TRN Transport Industries 11.GAS Gas 30.OME Machinery and Equipment 12.OMN Other Minerals 31.OMF Other Manufacturing 13.PCR Processed Rice 32.EGW Electricity, Water and Gas 14.MET Meat Products 33.CNS Construction 15.MIL Milk Products 34.T_T Trade and Transport 16.OFP Other Food Products 35.OSP Other Services (Private) 17.B_T Beverages and Tobacco 36.OSG Other Services (Govt) 18,TEX Textiles 37.DWE Ownership of Dwellings 19.WAP Wearing Apparels 18

19 Table 6: Percentage deviations in the average ad valorem tariff on iort by region NAM JPN AUS NZL CHINA SKOR TWN MYS/ SGP THA/ PHL IDN ROW t S T t Table 7: Values for coefficients and parameters from the GTAP database in the percentage-change form of the stylised model* NAM JPN AUS NZL CHINA SKOR TWN MYS/ SGP THA/ PHL IDN ROW S C S I S G S X S M (a) S L (b) S K S FC S T (c) σ M (d) σ X (e) σ KL * Source: GTAP database, version 5. (a) Ratio of the economy-wide cost of skilled and unskilled labour to the total cost of endowed commodities. (b) Ratio of the economy-wide cost of capital, land and natural resources to the total cost of endowed commodities. (c) Couted as a weighted average of the iort substitution elasticities across commodity in the home region. The weights were calculated from the cif-paid values of iorts. (d) Couted as a weighted average of the iort substitution elasticities by commodity across regions of destination. The weights were calculated from the fob values of exports. (e) Couted as a weighted average of the factor substitution elasticities across industries. There weights were based on total factor payments. 19

20 Table 8: Percentage-changes in key endogenous variables of the stylised model NAM JPN AUS NZL CHINA SKOR TWN MYS/SGP THA/PHL IDN ROW GTAP Stylised GTAP Stylised GTAP Stylised GTAP Stylised GTAP Stylised GTAP Stylised GTAP Stylised GTAP Stylised GTAP Stylised GTAP Stylised GTAP Stylised 1 y (r ) c i g x (a) m y fc (r ) k p gdp (r ) tot (a) p c (r ) p g (r ) rer p fc gdp (r ) rp l (r ) rp k (r ) rw (a) Exogenously set equal to the GTAP projections for exports via endogenous shifts in equation (7 ). (b) Exogenously set equal to the GTAP projections for the terms of trade via endogenous shifts in equation (11 ). 20